Machine Learning-Based Systemic Assessment of Political Instability Effects on Firm Performance

This study rigorously examines the impact of political instability (POI) on firm performance (FPER) using high-dimensional panel data from 2006 to 2022, drawn from the World Bank Enterprise Survey (WBES) for 60 economies. Using advanced machine learning-based econometric techniques, including Double-Selection LASSO Regression (DSLR) and Partialing-Out LASSO Regression (POLR), the analysis reveals that POI significantly reduces FPER across the sampled countries. These findings remain robust across a series of validation tests, including alternative estimation approaches such as Cross-Fit Partialing-Out LASSO Regression (CF-POLR) and Bayesian Model Averaging (BMA), the use of alternative FPER proxies—employment growth (FEMG), innovation (IINN), and labor productivity (FLP)—and the substitution of POI with government regulation (REG). Mediation analysis further indicates that operational costs (OCOST) and firm investment (FINV) significantly and partially mediate the total effect of POI on FPER. In contrast, financial constraints (FCST) do not emerge as a significant mediator. The moderation analysis shows that political connections (PC) substantially attenuate the negative impact of POI on FPER. Heterogeneity analyses demonstrate that small, young, and capital-intensive firms are more severely affected by POI than medium and large, older, technology-intensive, and labor-intensive firms. Additionally, firm ownership-based heterogeneity indicates that state-owned enterprises experience slightly stronger adverse effects from fluctuations in POI than non-state-owned firms. Based on these empirical insights, policymakers need to promote institutional stability and provide direct support for vulnerable young and small firms to reduce the adverse effects of POI on FPER. Ultimately, this boosts economic flexibility in politically unstable markets by managing key growth drivers.